Abstrakt
Protein-protein interactions (PPI) are crucial in a wide range of biological processes. Demands of their identification by experimental methods lead to development of computational methods.
However, their prediction quality is still far from perfect, thus we recently proposed INSPiRE - a new knowledge-based method for prediction of protein-protein interaction sites [1]. Another problem with available methods is that they are typically available as a web server only.
As a result, there is not possible a batch processing or it is limited by restrictions of individual web servers, and an integration of these methods into a pipeline is problematic. Also, there are no possibilities to modify individual methods and with only limited possibilities to parametrize them.
To avoid these deficiencies, we decided to make our method available as an open-source software. The software is written in C++ programming language to make it compilable on wide range of platforms and to reach a maximal efficiency.
Its back end is designed as a header-only to make its use in other people's projects as easy as possible. The back end is fragmented into independent modules that provide elementary operations (e.g. one module parses temperature of an amino acid, while another module computes its RASA value), so it is easy to write a new module (e.g. generating a new feature) and hook it up.
The software also has a front end that allows the user to promptly use INSPiRE. It allows the user to compile the whole pipeline into a single program, or compile each module as a separate program, which allows reusing of common features and/ or knowledge-bases and thus decrease time and space requirements during tuning of the pipeline.
In addition to the prediction of PPI, it is also possible to use the framework to distinguish between biological protein-protein interactions and crystal contacts, or to search homologous proteins or fragments. [1] Jelínek, J., Škoda, P., & Hoksza, D. (2017). Utilizing knowledge base of amino acids structural neighborhoods to predict protein-protein interaction sites.
BMC bioinformatics, 18(15), 492.